Loading ToolBench
An MCP server that rewrites AI-assisted prose, removes stock AI phrasing, and provides writing quality scoring with source preservation checks.
Zero Slop MCP exposes a single tool 'deslop' with a descriptive purpose statement but critical gaps in schema visibility and parameter documentation. The tool description is comprehensive (198 chars) and explains the use case well, meeting baseline standards. However, the input schema visible in the evaluation data shows three parameters (text, genre, audience) all typed as strings with descriptions, which is good, but there is NO visible output schema documentation in the provided source code. The tool description references 'writing scores', 'before and after' comparisons, and 'review warnings', but the actual response structure is not documented anywhere in the source excerpts. This violates the critical rule that 'Tools returning lists should accept page/offset and limit parameters and return a total count' and more fundamentally, 'Document the output schema' (pattern:tool, pattern:response-shaper). The tool is marked with risk='WRITE', indicating it modifies user content, but there is no confirmation/dry-run pattern visible. The parameter descriptions are adequate (40-60 chars each) but could be more actionable, e.g., the 'genre' and 'audience' parameters lack examples or clarification of expected values (enum vs free-form). No pagination, no per-item error handling, and no batching capability despite the name suggesting it might handle multiple rewrites. The server's .mcp.json and mcp.json files confirm HTTP (streamable-http) transport, which is current, but the tool definition itself lacks production-grade rigor.
Rewrite a pasted draft with one bounded AI editorial response plus local scoring and source checks. Returns the safest source-preserving edit and exact before and after writing scores. If a writing target is missed, the edit still comes back with a clear review warning. Use it to improve writing quality, never to hide authorship or evade a disclosure requirement. Try and MCP use our hosted Zero Slop agent harness; results and speed may differ across Codex, Claude Code, Cowork, ChatGPT Work, and other hosts or skills.
Output schema not documented. Tool description mentions 'writing scores', 'before and after comparison', 'review warning', but no response structure is visible in source code. LLMs cannot plan downstream tool calls or extract returned fields without knowing the response shape.
Parameter 'genre' and 'audience' lack enum constraints or format specification. Descriptions are generic ('Writing genre or style context', 'Target audience for the writing'). No examples or valid values listed. LLMs will guess at free-form strings, risking invalid API calls.
Tool modifies user content (risk=WRITE) but no dry-run, confirmation, or undo capability is documented. Per pattern:confirmation-request, irreversible operations should support a dry-run or confirmation step to prevent catastrophic agent errors.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | C | 67 | 2025-06-18+ | v2 |
No error recovery guidance. Tool description does not explain what happens on failure, how LLMs should retry, or what alternatives exist. Per pattern:recovery-guide, error responses must tell the LLM what to do next.
Tool name 'deslop' is not a verb-noun pattern (get_, create_, update_, etc.). While the meaning is contextually clear, it violates the critical naming rule that 'Tool names should start with a verb that reflects the action.' Rename to 'rewrite_text' or 'edit_prose' for clarity.